Article pubs.acs.org/OPRD
Selection and Scale-Up Evaluation of an Alternative Route to (−)-(3R,4R)‑1-Benzyl-4-(benzylamino)piperidin-3-ol Ian S. Young,* Adrian Ortiz, James R. Sawyer, David A. Conlon, Frederic G. Buono, Simon W. Leung, Justin L. Burt, and Eric W. Sortore† Chemical Development, Bristol-Myers Squibb Company, One Squibb Drive, New Brunswick, New Jersey 08903, United States ABSTRACT: An efficient, scalable synthesis of (−)-(3R,4R)-1-benzyl-4-(benzylamino)piperidin-3-ol (4) is described. Reduction of the pyridinium salt prepared from pyridine and benzyl chloride generated the corresponding tetrahydropyridine derivative. A two-stage epoxidation, followed by ring-opening of the epoxide with BnNH2, established the regiochemistry of the amino alcohol and served to set the trans-relationship between the amine and the hydroxyl group. The resulting racemic intermediate was then resolved by salt formation with (R)-O-acetyl mandelic acid. The process produced the O-acetyl mandelic acid salt of (−)-4 in 27% overall yield from benzyl chloride.
■
INTRODUCTION
and coupling partner 2 is generated from bis-benzyl derivative (−)-4. 3,4-Disubstituted piperidine (−)-4 is the compound of interest in this communication. The current large scale synthesis of piperidine (−)-4 features the ring expansion of pyrrolidinol 9, which is derived from (R)pyroglutamic acid (7).2 The overall process requires seven steps and proceeds in 22% overall yield (Figure 2). Although this strategy has been used to prepare greater than 200 kg of piperidine (−)-4, many of the transformations are operationally laborious. The preparation of (−)-4 requires significant plant time, leading to decreased throughput and increased production costs. Due to the need for large quantities (>100 kg) of 1, a more efficient synthesis of (−)-4 was desired. As previously reported,3 four alternative syntheses of compound (−)-4 were accomplished on lab scale (Figure 2). We began to evaluate the metrics/(dis)advantages of each strategy to select a route for further development. The initial sequence considered (route A, Figure 2) set the absolute and relative stereochemistry through asymmetric hydrogenation of enolized β-ketoester 10, and a Curtius rearrangement installed the second requisite nitrogen of 11. This route produced cyclic carbamate 11 in high overall yield (68%) and in short order (3 steps). Carbamate 11 has the potential to serve as an alternative protecting group to those currently used (Boc, Bn) during the coupling,4 although the API step deprotection would have to be redeveloped. Alternatively, 11 could be converted to 4 by a two step process (hydrolysis of carbamate and subsequent benzylation). While this work was being conducted, the cost and availability of the hydrogenation ligand 12 were of concern. Route B utilized material from the chiral pool and featured ring expansion5 of a 2-deoxy-D-ribose derived bis-amine 14. This strategy was shorter than the current route (5 vs 7 steps) but produced the desired target in only 15% overall yield (>99.9% ee). The key rearrangement to produce (−)-4 was low yielding and would require substantial optimization. The rearrangement was complicated by the observation that each
Lung cancer is responsible for over one-third of cancer related deaths in the United States, with the non-small cell variant being a particularly aggressive form. Treatments that prolong the life of those stricken with this disease are of the utmost importance and, as a result, are being actively investigated by the pharmaceutical industry. BMS-690514 (1) is a potential therapeutic agent for the treatment of lung and other cancers.1 From a retrosynthetic perspective, compound 1 can be dissected to protected piperidine 2 and pyrrolotriazine 3, the union of which is realized by an SN2 displacement of the activated primary alcohol (Figure 1). Compound 3 is derived from chloro-substituted pyrrolotriazine 6 and m-anisidine (5);
Figure 1. Retrosynthesis of BMS-690514 (1) to yield the intermediate of interest (4). © 2012 American Chemical Society
Received: June 29, 2012 Published: September 6, 2012 1558
dx.doi.org/10.1021/op300174w | Org. Process Res. Dev. 2012, 16, 1558−1565
Organic Process Research & Development
Article
Figure 2. Current route to piperidine (−)-4 and evaluation of alternative strategies. FGI = functional group interconversion; OAMA = (R)-O-acetyl mandelic acid.
albeit in racemic form (22).8 After opening with BnNH2, (±)-4 was produced in 50% yield (5 steps, unoptimized) from inexpensive and readily available benzyl chloride (19) and pyridine (20).9 After resolution with (R)-O-acetyl mandelic acid (OAMA), the salt of 4 was obtained in 22% overall yield (93.8% ee after salt break).10 Upon evaluating the scalability of this sequence, we envisioned a reduction in the number of operations through telescoping transformations. Challenges associated with this route were that some intermediates were identified to be oils and salt formation was necessary for crystallization. Additionally, further resolution development would be required to increase enantiomeric excess. With four sequences available, we defined our criteria for route selection: (a) a reduction in step count compared to the current seven steps; (b) utilization of inexpensive starting materials; and (c) a reduction in the total number of operations. Based on the summaries in Figure 2, we deemed that Route D had the greatest potential to meet these requirements. Only five steps were required from inexpensive pyridine and benzyl chloride, and the reactions on lab scale were operationally simple to perform. Areas requiring further study were identified during our lab scale campaign (lack of
anomer of 14 displayed differing behavior upon subjection to catalytic hydrogenolysis. One anomer converted to the desired product (4) more rapidly (1−2 h), and if this anomer was subjected to the extended reaction times (4.5 h) required to convert the other anomer to 4, substantial product Ndebenzylation occurred. To overcome this, the anomers had to be separated prior to rearrangement. Another concern was the necessity to purify the polar intermediates by chromatography, as anomeric mixtures at each stage rendered isolation by crystallization difficult. Route C also utilized 2-deoxy-D-ribose (13) as a feedstock, from which a bis-mesylate intermediate was prepared. Double displacement of the mesylates with BnNH2 induced closure to piperidine 17. Further functional group manipulations6 installed the epoxide of 18, which was regioselectively opened (>20:1) with BnNH2 in the presence of a lithium salt.7 Overall, seven steps were required to produce (−)-4 in 10% yield (>99.9% ee), with inefficient epoxide formation (26%) being the major contributor to the low overall yield. Instability of the bis-mesylate intermediate and difficulties encountered during purifications were also areas of concern. The final sequence considered (Route D) featured an alternative synthesis of the epoxide (18) used in Route C, 1559
dx.doi.org/10.1021/op300174w | Org. Process Res. Dev. 2012, 16, 1558−1565
Organic Process Research & Development
Article
crystalline/isolable intermediates, and the requirement for resolution optimization). Preliminary results during the lab scale development of this route [fumaric acid salt of epoxide 22 for isolation, and (R)-O-acetyl mandelic acid for the resolution of (±)-4 (44%, 93.8% ee)]10 served as a starting point for addressing these issues. The following reports our findings and optimizations upon performing this chemistry on scales approaching 1 kg.
Scheme 2. Addition Modes Evaluated by Calorimetry for the Reduction of Pyridinium Salt 23 and the 20 L Procedurea
■
RESULTS AND DISCUSSION11 Although not discussed in our original communication,3 tetrahydropyridine 21 was prepared by the alkylation of pyridine with benzyl chloride and subsequent reduction.12 The initial lab-scale procedure required slow addition of benzyl chloride to neat pyridine at 120 °C to prevent a delayed exotherm (Scheme 1). The resulting salt 23 was dissolved in
For A and B, values are for addition of 0.1 equiv of limiting reagent (NaBH4 and 23, respectively).
a
3). In the reverse mode of addition [23 (0.25 mmol in ethanol) to an ethanol solution of NaBH4 (2.5 mmol) (mode B, Scheme
Scheme 1. Preparation of Tetrahydropyridine 21 by Original Lab Scale (o) and Proposed Modified (m) 20 L Procedures
methanol, and solid sodium borohydride was added portionwise to prevent excessive foaming. After workup, tetrahydropyridine 21 was isolated by vacuum distillation. As a means to alleviate the first issue (neat reaction conditions, 120 °C), we evaluated the performance of this alkylation in toluene. Although the precipitated product could be isolated by filtration, the hygroscopicity of 23 limited the practicality of this approach. We then considered the possibility of performing the alkylation reaction in ethanol, as this would allow the reduction to be performed in the same reactor without isolation of salt 23. Experimentally, the reaction required 8−10 h at reflux for complete consumption of benzyl chloride. A slight excess of pyridine (7%) was utilized, as the presence of residual benzyl chloride during the subsequent reduction led to quaternization of reduction product 21. With an ethanolic solution of 23 in hand, we began development of a scalable sodium borohydride reduction. During the original lab scale developments, solid sodium borohydride (1.2 equiv) was added portionwise to a methanol solution of salt 23 to minimize foaming/exotherm. It was estimated that this addition would take 3−4 h on 20 L scale, rendering this addition method impractical. To alleviate the issue, we considered two protocols for reduction: 1) addition of a basic, aqueous solution of sodium borohydride to the alcoholic solution of 23; and 2) addition of the alcoholic solution of 23 to a slurry of sodium borohydride in ethanol. It was presumed that each of these alternatives would yield a dose controlled reaction rate. Prior to scaling this reaction to 20 L, we initiated a study of the reduction by calorimetry to identify potential thermal events. Initially, the mode of addition (NaBH4 to 23 vs 23 to NaBH4) was evaluated on the basis of heat evolution. Charging an ethanol solution of sodium borohydride (0.25 mmol, 10 mol %) to a solution of 23 (2.5 mmol) in ethanol at 10 °C (mode A, Scheme 2) generated 79 J over a 6 min reaction time (Figure
Figure 3. Heat flow versus time detected by isothermal reaction calorimetry.
2)], the total heat detected was 464 J, with the exotherm being self-sustaining for more than 140 min. There was also a substantial increase in pressure within the calorimeter during this mode of addition. The marked difference in heat evolution (79 vs 464 J) resulted from the tertiary amine reduction product13 (21) catalyzing the exothermic decomposition14 of the remaining sodium borohydride (0.9 equiv). To validate this theory, addition of a catalytic amount of tetrahydroypyridine 21 to a slurry of sodium borohydride in ethanol initiated violent gas evolution, indicating that reverse addition of pyridinium salt 23 to NaBH4 was not feasible. Further lab scale optimization in conjunction with calorimetry studies identified the addition of a solution of NaBH4 (1.15 equiv) in aqueous NaOH (0.01 M)15 to an ethanol solution of 23 as a viable procedure for scale-up (Scheme 2, 20 L procedure). For this process, the heat detected by reaction calorimetry was 80 J and the adiabatic temperature rise was estimated to be 104 °C. Upon implementation on 20 L scale, addition of the NaBH4 solution over 4 h led to minimal exotherm and allowed the reaction temperature to be maintained below 20 °C. With a process to safely produce tetrahydropyridine 21 on scale, we turned our attention to improving the epoxide formation sequence. For lab scale development, distilled 21 was treated with NCS in the presence of TFA to form the 1560
dx.doi.org/10.1021/op300174w | Org. Process Res. Dev. 2012, 16, 1558−1565
Organic Process Research & Development
Article
upon direct addition of fumaric acid to the refluxing epoxide solution also needed to be addressed. Due to the insolubility of fumaric acid in most organic solvents, slow addition of a fumaric acid solution to the epoxide stream was not feasible. After exploring a variety of mixed solvent systems, it was found that fumaric acid was soluble in 11 volumes of 10:1 acetone/ water at reflux (higher ratios of water further increased fumaric acid solubility but led to diminished recovery of the salt 26). After reducing the temperature of the fumaric acid solution to 45 °C (the minimum temperature at which the fumaric acid would remain soluble), seeds of 26 (10 mol %) were added. The crude epoxide solution in 2-MeTHF (used instead of MTBE for extraction of the epoxide) was then dosed at a rate to control the crystallization. Salt 26 was obtained in 59% yield (from benzyl chloride) with a purity of 94−96 LCAP (liquid chromatography area percent, crude LCAP = 76−80). The addition temperature of 45 °C was a key process parameter. If the epoxide solution is added to the fumaric acid solution at temperatures > 50 °C, a significant amount (up to 12%) of byproduct (27) from epoxide opening by fumaric acid is observed in the isolated product. With the fumaric acid salt (26) in hand we were prepared to evaluate epoxide opening and resolution on scale. For the lab scale synthesis, fumaric acid was removed by basic washes to yield epoxide 22, and then epoxide opening (LiCl, BnNH2, CH3CN) proceeded with a high degree of regioselectivity (>20:1)7c to yield (±)-4 (Scheme 5). Racemic 4 was
chlorohydrins 24/25 (Scheme 3). The crude product was then isolated and subjected to potassium carbonate (2.0 equiv) in Scheme 3. Preparation of the Racemic Epoxide by Both Original Lab Scale (o) and Modified 20 L (m) Methods
methanol to yield epoxide 22. For the 20 L campaign, the crude tetrahydropyridine stream from reduction of pyridinium 23 was used directly. A similar protocol for epoxide formation was utilized, with the exception of the direct addition of 10 N NaOH to the reactor once chlorohydrins formation was complete. This removed the necessity for workup and isolation of 24/25. It was necessary to charge 4.2 equiv of sodium hydroxide to compensate for the acidic components [TFA (1.25 equiv) and succinimide (1.2 equiv), leaving 1.75 equiv of NaOH for reaction] present from chlorohydrin preparation. After three telescoped steps, the crude epoxide 22 was extracted into an organic solvent in preparation for salt formation/ crystallization with fumaric acid. On lab scale, crude epoxide 22 was extracted into MTBE, concentrated, and then redissolved in acetone. After heating this solution to reflux, solid fumaric acid (1.0 equiv) was added directly to the flask (Scheme 4). This led to rapid precipitation
Scheme 5. Preparation of Racemic 4 and Resolution by Both Original Lab Scale (o) and Modified 20 L (m) Methodsa
Scheme 4. Preparation of Fumaric Acid Epoxide Salt 26 by Both Original Lab Scale (o) and Modified 20 L (m) Methodsa a
a
FA = fumaric acid; OAMA = (R)-O-acetyl mandelic acid.
crystallized from toluene/heptane (85%) and then subjected to the unoptimized resolution conditions [ethanol (anhydrous), (R)-O-acetyl mandelic acid] to yield 28 (42%, 93.8% ee).10 Upon scaling this sequence to 20 L, our objectives were to eliminate the isolation/crystallization of racemic 4 prior to resolution, and improve both the recovery and enantiomeric excess of isolated 28. The epoxide opening conditions (LiCl, BnNH2, CH3CN) used on lab scale to generate racemic 4 yielded similar results on scales approaching 1 kg and were not modified. Further development of the isolation/resolution of racemic 4 supported the continued use of (R)-O-acetyl mandelic acid (29) in alcoholic solvents. Utilizing previously crystallized (±)-4, the greatest enantioenrichment (97.3% ee)10 was achieved by salt formation in EtOH [SDA-3A, 190 proof (the SDA-3A designation indicates denaturing with 5 parts methanol)], but the recovery was modest (36%). This reduction in yield is presumably due to water present in the ethanol leading to increased solubility of the product salt (Figure 4). The use of anhydrous solvents such as ethanol (200 proof) and n-BuOH
LCAP = liquid chromatography area percent.
of salt 26, and the addition of solid to the refluxing solution caused excessive foaming. Due to the uncontrolled crystallization, excess fumaric acid contaminated the isolated product. Nonetheless, this procedure offered the first isolable, solid intermediate in this sequence. Modifications were necessary to make this salt formation amenable to 20 L scale, as better control of both crystallization rate and purity upgrade were required. The foaming observed 1561
dx.doi.org/10.1021/op300174w | Org. Process Res. Dev. 2012, 16, 1558−1565
Organic Process Research & Development
Article
Figure 4. Optimization of the (R)-O-acetyl mandelic acid resolution of racemic 4.
was examined to circumvent this issue, and while both solvents yielded comparable enantiomeric excess (93.8 and 93.5%, respectively), the recovery with n-BuOH was superior (47 vs 42%). Further improvements with n-BuOH were realized by seeding the crystallization (1% loading) and utilizing a temperature cycle.16 This process allowed (−)-4 to be isolated in 94.6% enantiomeric excess and 44% recovery from previously isolated racemic 4. With an optimal resolution developed using previously crystallized (±)-4, we attempted to use the (R)-O-acetyl mandelic acid crystallization as a means of both resolution and purification. Upon completion of epoxide opening by BnNH2, water was added to the reaction mixture and crude (±)-4 was extracted into n-BuOH. This aqueous extraction was necessary to remove lithium chloride (its presence leads to uncontrolled precipitation of the racemate during the salt formation), and upon addition of O-acetyl mandelic acid, 28 was produced in 43% yield and 96.0% ee (Scheme 5).10 The chemical purity of (−)-4 that was isolated after salt break was greater than 99.9 LCAP, indicating that the crystallization effectively rejected the impurities from the epoxide opening sequence. The enantiomeric excess garnered from the resolution was still insufficient for our purposes; however, the route to coupling partner 2 offered three opportunities for enantioenrichment. Initially we examined crystallization of the material resulting from salt break of 28 (option 1, Figure 5). It was found that either acceptable enantioenrichment or recovery could be achieved by crystallization from varying ratios of toluene/heptane, but both could not be realized simultaneously. The second option (Figure 5) involved conversion of salt 28 to the coupling partner 2 and then crystallization from toluene/heptane. Unfortunately, a similar outcome to the crystallization of (−)-4 (either good recovery or chiral purity, but not both) was found. The final option investigated (option 3, Figure 5) involved simple recrystallization of salt 28 resulting from the resolution. Since we were now separating diastereomeric salts rather than enantiomers, this enantioenrichment strategy proved to be the most effective. A range of alcoholic solvents were surveyed, and recrystallization from ethanol (SDA-3A 190 proof, 10 vol) yielded 28 in 99.9% ee with an
Figure 5. Strategies investigated for enantioenrichment.
acceptable 93% recovery.10 With an enantioenrichment strategy in place, it appeared that we could efficiently execute the sequence on 20 L scale. This material (28) could then be converted to coupling partner 2 using the current chemistry or be used to explore alternative protecting strategies for coupling. However, at this point we obtained occupational toxicology data that necessitated modification to our isolation strategy. A local lymph node assay (LLNA) identified the epoxide fumaric acid salt (26) as a dermal sensitizer.17 This introduced material handling restrictions, and we sought to eliminate the formation/isolation of the fumaric acid salt entirely. This would stress our resolution crystallization, which would now afford the only opportunity for purification from benzyl chloride/pyridine. Progression through the sequence yielded epoxide 22, which was not isolated, but instead carried directly into the epoxide opening process. Reaction of crude 22 with BnNH2 behaved similarly to when isolated epoxide salt 26 was used, and with the dark brown stream of (±)-4 in hand (60 LCAP), we were now ready to test the limits of our resolution in terms of impurity purge. Using the previously developed conditions (nBuOH, 1% seeding, 1 temperature cycle), target salt 28 was isolated as a white crystalline solid (29% yield from BnCl; 97 LCAP, 95.0% ee) on a 20 L scale. This material was then recrystallized from ethanol (SDA-3A, 190 proof) for 1562
dx.doi.org/10.1021/op300174w | Org. Process Res. Dev. 2012, 16, 1558−1565
Organic Process Research & Development
Article
generation 20 L route required five steps and three crystallizations to obtain O-acetyl mandelic salt 28 in 23% overall yield (100 LCAP and 99.9% ee). At this point, epoxide salt 26 was identified as a dermal sensitizer, and the process was modified to eliminate formation/isolation of this compound. This further streamlined the route to 28 and allowed us to arrive at our current process. It was estimated that this route would offer a cost reduction of 60% when compared to the current strategy utilizing pyroglutamic acid [for the multikilogram production of intermediate (−)-4]. This potential savings is remarkable, when one considers that more than half of the material is removed during the resolution. A full account reporting the development work toward the preparation of BMS-690514 (1) on scale will be forthcoming.
enantioenrichment and chemical purification (94% recovery, 100 LCAP, 99.8% ee) to produce 28 in 27% overall yield from benzyl chloride (Scheme 6). More importantly, this outcome demonstrated that formation/isolation of dermal sensitizer 26 is not necessary. Scheme 6. Optimized Sequence to 28a
■
a
EXPERIMENTAL SECTION General. Reactions were performed under an atmosphere of nitrogen unless otherwise noted. Reagents were used as received unless otherwise noted. Reported yields are for isolated materials or calculated solution yields and are corrected for potency. NMR spectra were recorded on a Bruker DRX-500 instrument and are referenced to residual undeuterated solvent. The following abbreviations are used to explain multiplicities: br = broad, s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet. Infrared (IR) spectra were recorded on a PerkinElmer 1600 series FT-IR spectrometer. High-resolution mass spectra (HRMS) were recorded on a Thermo Orbi-trap Discovery instrument. Reaction calorimetry was studied using an Omnical Insight RT-10 isothermal reaction calorimeter. Synthesis of Tetrahydropyridine 21. Caution: Compound 21 was found to catalyze the exothermic decomposition of sodium borohydride. As a result, sodium borohydride should be added to a solution of the pyridinium salt, instead of adding the pyridinium salt solution to a slurry of sodium borohydride. For further discussion, see text. EtOH (SDA-3A, anhydrous, 2 L) was charged to a 20 L glass reactor equipped with a reflux condenser. Benzyl chloride (800 g, 6.32 mol, 1.0 equiv) and pyridine (533 g, 6.74 mol, 1.07 equiv) were charged into the reactor, and the heating jacket was set to 85 °C. The reaction was refluxed until HPLC analysis (210 nm) indicated that benzyl chloride had been consumed [
